Development of a Hybrid Metamodel based Simulation Optimization Algorithm
In this paper, a metamodel based hybrid algorithm was developed for optimization of digital computer simulation models. The simulation models are considered to be computationally expensive. It is also considered to have a single stochastic and unconstrained response function. The hybrid algorithm is developed by modification and integration of several concepts and routines. We employed the nested portioning and the particle swarm optimization algori-thms to develop an efficient search mechanism for the hybrid algorithm. Then we integrated the modified Kriging metamodel to the search mechanism for facilitating the function fitting processes of the simulation’s output. The efficiency of the developed hybrid algorithm was then evaluated through computational experiments. Ten complex test problems were selected from the literatures and the efficiency of the developed hybrid algorithm was evaluated by comparing its performances against three known algorithm which are cited in the literature. The result of these computational experiments revealed that the developed hybrid algorithm can provide very robust solutions with a very low computational effort.
